Director Computer

SECURITY & PRIVACY ECONOMICS

Editors: Michael Lesk, ****@***.*** Jeffrey MacKie-Mason, ***@*****.***

Can Users Control Online Behavioral

Advertising Effectively?

Lorrie Faith Cranor Carnegie Mellon University

O nline behavioral advertising companies typically place an opt-

(OBA) is the increasingly out cookie on the user s computer

w idespread practice of targeting i nstead of the tracking cookie that

users with specific online ads on u niquely identifies the user. Com-

the basis of a user s previous online panies can continue to track users,

behavior. Advertisers pay a pre- but they re prohibited from deliv-

mium for targeted ads because users ering targeted ads to users who

are more likely to make purchases have opted out.

a fter viewing relevant ads.1 On the Web browser vendors have also

other hand, whereas some users taken steps to help users opt out.

might appreciate seeing more rele- A ll major browsers let users selec-

vant advertisements, many say they tively block cookies, which can

find targeted advertising creepy effectively reduce tracking. In addi-

and don t like the idea of companies tion, Microsoft lets users install

tracking their online activities.2,3 t racking protection lists (TPLs) in

Many tools empower users to con- I nternet Explorer 9, which specify

trol whether and when they re domains from which Web requests

tracked for behavioral advertising; should be blocked as well as excep-

however, whether users can effec- tions to blocking rules. Users who

tively control tracking and OBA want to avoid OBA can download a

using these tools is unclear. T PL that blocks requests to known

t racking companies. These are

Current Efforts available from several organiza-

The US Federal Trade Commis- tions, including TRUSTe, Privacy-

sion has pressured companies to Choice, and Abine.

a llow users to easily opt out of In February 2012, in con-

OBA.4,5 I n response, the Digi- junction with the White House

tal Advertising Alliance (DAA) announcement of a Consumer

developed self-regulatory guide- Privacy Bill of Rights, the DAA

lines that require companies to announced plans to include

notify users about behavioral browser-based choice mechanisms,

advertising and allow them to opt such as Do Not Track, as part of its

out.6 The DAA created a standard- self-regulatory program. Inter-

ized OBA icon for companies to net Explorer and Firefox also

place on their behavioral adver- have a setting that sends a Do Not

tisements along with the clickable Track header with every HTTP

tagline AdChoices. It also offers request, and the World Wide

a website (www.aboutads.info/ Web Consortium has convened a

choices) where users can opt out of working group to develop its speci-

targeted ads from dozens of com- fications. In the meantime, there s

panies. When a user opts out, ad no standard interpretation of what

93

1540-7993/12/$31.00 2012 IEEE Copublished by the IEEE Computer and Reliability Societies March/April 2012

SECURITY & PRIVACY ECONOMICS

activities the word track denotes, participants information on an opt- Most participants said they were

a nd most companies ignore Do out tool and asked them to install it w illing to allow targeted advertis-

Not Track headers. on our laboratory computer. They ing while they conducted some

Several browser add-ons help asked participants to configure the t ypes of searches, but not others.

users avoid tracking. These tools tool according to their personal pref- In addition, they tended to be most

can set opt-out cookies, block erences, then describe the configu- comfortable being targeted by ad

advertising cookies, block requests rations they chose. Then they asked companies whose names they were

to tracking domains, or notify participants to configure the tools familiar with, and most wary of ad

users about which trackers are to match a set of provided specifi- companies they d never heard of.

present on the websites they visit. cations. Finally, they asked partici- When shown the industry

The online-advertising indus- pants to perform several browsing advertising icons out of context, 41

try has trumpeted these efforts tasks using the tools configured out of 48 participants didn t recog-

to demonstrate to regulators that w ith fairly protective settings. Some nize them. When shown the icons

t hey can effectively self-regulate. of these tasks required third-party in context next to an advertisement

A lthough these tools might be a content, cookies, or scripts to func- w ith the accompanying tagline,

good first step, we can t conclude tion properly and thus couldn t be most participants didn t recognize

t hat self-regulation is effective completed when some of the tools them and couldn t figure out what

w ithout empirical data on whether were set to block tracking. Partici- they indicated. Five participants

users who want to limit OBA are pants were advised that they could realized that the icons indicated

able to use the tools effectively. adjust the tools settings if needed to that ads were being tailored, but

complete the tasks. none of them understood that the

User Studies icons were also supposed to inform

User Knowledge

A series of studies at the CyLab them that data was being collected

and Perceptions

Usable Privacy and Security for targeting. Some participants

(CUPS) Laboratory at Carnegie When asked about the first thing thought the icons were intended for

Mellon University assessed the t hat comes to mind when they hear people who wanted to advertise on

effectiveness of tools to limit OBA. Internet advertising, many par- websites. Many didn t expect the

These include studies of users per- ticipants mentioned popups, said icons to be clickable or had miscon-

ceptions of OBA and users ability t hey found Internet advertising ceptions about what would happen

to use opt-out tools.7 a nnoying, or said they routinely if you clicked them. Some feared

Last summer, our research team, ignored it. However, when asked that clicking the icon would lead to

which included my students Pedro whether Internet advertising was more ads or popups. These results

Leon and Blase Ur, recruited 48 useful, more than half the partici- suggest that the icons and tagline

Internet Explorer 9 and Firefox pants said yes, but many wished it are failing to effectively communi-

5 users from the Pittsburg area was less obtrusive. Most partici- cate their purpose to users.

to our laboratory for individual pants also said that receiving ads Before discussing opt-out tools,

semistructured interviews and an tailored to their interests was use- L eon and Ur asked participants,

opportunity to learn more about ful. Before watching the video, few Are you aware of any ways that

online privacy tools. They screened participants knew how ads were can help you stop receiving tar-

out participants with computer tailored, and some had miscon- geted ads? About half the partici-

science or IT degrees or jobs. The ceptions about how OBA worked. pants mentioned deleting cookies,

interviews each lasted approxi- A fter watching the video, most something the video mentioned.

mately 90 minutes. Interview ses- participants were upset that com- Twelve participants didn t think

sions began with questions focusing panies track them without their t hey could do anything to stop

on participants prior knowledge k nowledge and consent and said receiving targeted ads. None men-

and attitudes about OBA and indus- t hey wanted to be able to control tioned opt-out cookies, industry

try advertising icons. Participants OBA. Several described OBA as opt-out websites, Do Not Track, or

v iewed an informational video on scary or creepy. Their miscon- T PLs, providing further evidence

OBA produced by the Wall Street ceptions were often fueled by their t hat consumer awareness of opt-

Journal, then discussed their under- d islike of popup ads and mistrust out options is fairly low.

standing of and attitudes about of the advertising industry. Some

Opt-Out Tool Evaluation

OBA and several online advertis- were concerned that their contact

ing companies. i nformation and financial records Successful use of opt-out tools

Then Leon and Ur gave m ight be collected during OBA. requires that users can install a

94 IEEE Security & Privacy March/April 2012

tool, configure it to match their doesn t guide users to subscribe to mechanisms also provided no feed-

preferences, and use the tool effec- a T PL, which is necessary for the back, and there s currently no way

tively. Leon and Ur tested the T PL feature to provide protection. for tools to confirm that Do Not

usability of nine representative Furthermore, if users proactively Track preferences are being hon-

tools from three broad categories download a browser add-on, such ored. In contrast, for every website

for controlling OBA: as Ghostery or TACO, or visit users visited, Ghostery and TACO

a n opt-out website, they likely displayed notifications about which

three tools that set opt-out i ntend to block tracking. However, companies were attempting to track

cookies the DAA opt-out web- Ghostery and TACO don t auto- them and whether trackers had been

site, a similar website hosted by matically block any trackers. blocked. Users appreciated this

Evidon that includes opt-outs feedback and gained an understand-

Communication Problems

f rom more companies, and the ing of what the tools were doing.

PrivacyMark bookmark tool that Overall, the tools were ineffective

Users Want Protections

sets opt-out cookies for more at communicating their purposes

That Don t Break Websites

t han 160 companies whenever and guiding users to properly con-

it s clicked; figure them. They tended to pre- Participants had difficulty deter-

two built-in browser settings sent information at a level that mining when the tools they were

Internet Explorer 9 and Firefox was either too simplistic to inform using caused parts of websites to

5; and users decisions or too technical to stop working. In cases in which

four blocking tools Ghostery, be understood. For instance, Inter- some content wasn t displayed or

TACO (Targeting Advertising net Explorer 9 provides a simplistic features stopped working, partici-

Cookie Opt-Out), Adblock Plus, privacy slider whose six levels (for pants believed that their Internet

a nd Internet Explorer Tracking instance, medium ) don t describe connection was the problem. TPLs

Protection. their functionality. In contrast, have the potential to address this

participants couldn t understand problem by letting users subscribe

None of these nine tools empow- the jargon-filled technical explana- to a curated list that blocks most

ered study participants to effectively tions next to the slider. Ghostery t rackers except those that are nec-

control tracking and behavioral and TACO used terms that were essary for sites to function. How-

advertising according to their per- meaningless to participants: Web ever, participants were unaware

sonal preferences. tracker, Web bug, Flash cookie, t hat they needed to select a TPL or

Silverlight cookie, tracking u nsure how to decide which TPL

Users Couldn t Distinguish cookie, script, IFrame, and to select. In addition, sites may

between Trackers targeted ad network. In addition, bundle essential functionality into

The opt-out websites and the participants testing opt-out tools t rackers to prevent their trackers

Ghostery and TACO browser add- didn t understand what the tools f rom getting blocked.

ons provided users with lists of would opt them out of. They often

Confusing Interfaces

companies that they can block or mistakenly believed that they were

f rom which they can opt out. How- protected against tracking when Most tools suffered from major

ever, participants didn t recognize they were still being tracked even usability flaws. For instance, mul-

t he majority of these companies though they no longer saw targeted tiple participants opted out of only

a nd generally chose the same set- ads. Furthermore, users thought one company on the DAA s web-

tings for all companies on the list. deleting their cookies would site, despite intending to opt out

They couldn t set opt-out or block- increase their privacy, not realizing of all. Others mistook the page on

ing preferences meaningfully on a that deleting their cookies would which advertising companies reg-

per-company basis. a lso delete opt-out cookies (thus ister for the DAA for an opt-out

undoing their opt-out). page. Participants testing TACO

Inappropriate Defaults never realized that they weren t

Need for Feedback

The default settings for most of blocking any trackers. Participants

t he tools weren t appropriate for Many of the tools provided insuf- d idn t understand Adblock Plus s

users interested in protecting their ficient feedback. Participants were fi ltering rules. None of the partici-

privacy. Once a user enables a pri- unsure of what opting out meant pants who tested Internet Explorer

vacy feature, a protective default and how they could tell whether Tracking Protection realized that

for that feature seems reason- the opt out or cookie block- t hey needed to subscribe to TPLs

able. However, Internet Explorer ing was working. Do Not Track u ntil prompted in a later task.

95

www.computer.org/security

SECURITY & PRIVACY ECONOMICS

More emphasis on tool usability advertising works and are unfamil- Beliefs and Behaviors: Internet

is necessary to empower users to iar with online-advertising com- Users Understanding of Behavioral

control behavioral advertising. panies. When faced with choices Advertising, Proc. Telecommunica-

about blocking trackers from doz- tions Policy Research Conf. (TPRC

ens of unfamiliar companies, users 10), 2010; http://ssrn.com/abstract

T hese studies show that users can t make informed decisions. We =1989092.

lack awareness of the tools shouldn t expect users to read doz- 3. J. Turow et al., Americans

they can use to control targeted ens of privacy policies8 or become R eject Tailored Advertising and

advertising and the ability to use privacy experts. Tools that let users Th ree Activities That Enable It,

them effectively. Although the make coarse-grained choices and 2 009; http://ssrn.com/abstract

industry has developed guidelines translate these into the appropri- =1478214.

and an opt-out program, users ate fine-grained settings (perhaps 4. Protecting Consumer Privacy

either don t recognize the opt-out learning from other users or from i n an Era of Rapid Change, Fed-

icon or don t realize they can click users preferences or behaviors over eral Trade Commission, 2010;

it to get relevant information. In time) might offer a possible solu- w ww.ftc.gov/os/2010/12/101201

January 2011, the DAA announced tion. Privacy regulations that pro- privacyreport.pdf.

it was launching a marketing vide a baseline level of protection 5. J. Leibowitz, Concurring State-

campaign and website (www. offer a complementary solution that ment of Commissioner Jon

youradchoices.com) to inform con- might address user concerns. L eibowitz: FTC Staff Report: Self-

sumers about the AdChoices icon. Regulatory Principles for Online

References

W hether this campaign increases Behavioral Advertising, Feb.

awareness remains to be seen. 1. H . Beales, The Value of Behav- 2 009; www.ftc.gov/os/2009/02/

Most of the tools examined could ioral Targeting, Network Adver- P085400behavadleibowitz.pdf.

be substantially improved with tising Initiative, Jan. 2010; www. 6. Self-Regulatory Principles for

more attention to usability. How- net workadvertising.org/pdfs/ Online Behavioral Advertis-

ever, an underlying challenge is that Beales_NAI_Study.pdf. ing, Digital Advertising Alli-

users don t understand how online 2. A .M. McDonald and L.F. Cranor, ance, July 2009; www.aboutads.

info/resource/download/seven

-principles-07-01-09.pdf.

7. P.G. Leon et al., Why Johnny

Can t Opt-Out: A Usability Eval-

uation of Tools to Limit Online

PURPOSE: The IEEE Computer Society is the world s largest association of computing professionals and is the

Behavioral Advertising, Proc. Conf.

leading provider of technical information in the field. Visit our website at www.computer.org.

OMBUDSMAN: Email ****@********.***.

Human Factors in Computing Systems

Next Board Meeting: 11 15 June, Seattle, COMPUTER SOCIETY OFFICES

(CHI 12), ACM, 2012; www.cylab.

Wash., USA Washington, D.C.: 2001 L St., Ste. 700, Washington,

D.C. 20036-4928

cmu.edu_cylab11017.html.

EXECUTIVE COMMITTEE Phone: +1-202-***-**** Fax: +1-202-***-****

Email: **.***@********.***

8. A .M. McDonald and L.F.

President: John W. Walz*

Los Alamitos: 10662 Los Vaqueros Circle, Los Alamitos,

President-Elect: David Alan Grier;* Past President:

C A 90720-1314 Phone: +1-714-***-**** Email:

C ranor, The Cost of Reading

Sorel Reisman;* VP, Standards Activities: Charlene

****@********.***

(Chuck) Walrad; Secretary: Andre Ivanov (2nd VP);* VP,

P rivacy Policies, I /S: J. Law and

Membership & Publication Orders

Educational Activities: Elizabeth L. Burd;* VP, Member

Phone: +1-800-***-**** Fax: +1-714-***-****

& Geographic Activities: Sattupathuv Sankaran; VP,

Policy for the Information Society,

Email: ****@********.***

Publications: Tom M. Conte (1st VP);* VP, Professional

Asia/Pacific: Watanabe Building, 1-4-2 Minami-

Activities: Paul K. Joannou;* VP, Technical & Conference

2 008 Privacy Year in Review

Aoyama, Minato-ku, Tokyo 107-0062, Japan Phone:

Activities: Paul R. Croll; Treasurer: James W. Moore,

+81 3 3408 3118 Fax: +81 3 3408 3553 Email:

CSDP;* 2011 2012 IEEE Division VIII Director: Susan

i ssue; http://lorrie.cranor.org/

*****.***@********.***

K. (Kathy) Land, CSDP; 2012 2013 IEEE Division V

Director: James W. Moore, CSDP; 2012 IEEE Division

pu b s/re a d i n g Pol ic y C o s t

IEEE OFFICERS

Director VIII Director-Elect: Roger U. Fujii

President: Gordon W. Day; President-Elect: Peter

*voting member, nonvoting member of the Board of Governors

- authorDraft.pdf.

W. Staecker; Past President: Moshe Kam; Secretary:

Celia L. Desmond; Treasurer: Harold L. Flescher;

BOARD OF GOVERNORS

President, Standards Association Board of Governors:

Term Expiring 2012: Elizabeth L. Burd, Thomas M.

Steven M. Mills; VP, Educational Activities: Michael R.

Lorrie Faith Cranor is an associate

Conte, Frank E. Ferrante, Jean-Luc Gaudiot, Paul K.

Lightner ; VP, Membership & Geographic Activities:

Joannou, Luis Kun, James W. Moore, William (Bill) Pitts

Howard E. Michel; VP, Publication Services &

professor of computer science and

Term Expiring 2013: Pierre Bourque, Dennis J. Frailey,

Products: David A. Hodges; VP, Technical Activities:

Atsuhiro Goto, Andr Ivanov, Dejan S. Milojicic, Paolo

of engineering and public policy

Frederick C. Mintzer; IEEE Division V Director: James

Montuschi, Jane Chu Prey, Charlene (Chuck) Walrad

W. Moore, CSDP; IEEE Division VIII Director: Susan K.

at Carnegie Mellon University.

(Kathy) Land, CSDP; IEEE Division VIII Director-Elect:

EXECUTIVE STAFF

Roger U. Fujii; President, IEEE-USA: James M. Howard

Executive Director: A ngela R. Burgess; A ssociate

Contact her at ******@***.***.

Executive Director, Director, Governance: A nne

Marie Kelly; Director, Finance & Accounting: John

Miller; Director, Information Technology & Services:

Ray Kahn; Director, Membership Development:

Selected CS articles and columns

V iolet S. Doan; Director, Products & Services: Evan

Butterfield; Director, Sales & Marketing: Chris Jensen

are also available for free at

revised 22 Feb. 2012

http://ComputingNow.computer.org.

96 IEEE Security & Privacy March/April 2012

Contact this candidate